Send MIDI Sequencer to a MIDI Output

[novaXire]

Hello,

I would like to know if you have plan to add the possibility to send MIDI information to a MIDI Output or if you want to lock alphaTab to use only TinySoundFont?

I made a new class in AlphaTab.Windows and replace _synthesizer in alphaTab by this new class and MIDI works.

[Danielku15]

Hi.
AlphaTab is not locked to use only TinySoundFont, also not today. But I also think that the underlying synthesizer components are not suitable to send midi directly to another device. AlphaTab does not have a real time synthesizer or a real time sequencer. It does not send the midi events exactly at the time when they should be played. Instead it internally does the synthesizing always as fast as possible and buffers the samples until some extend.

The logic internally works like this: On each platform there is a raw PCM audio output (Web Audio Context, NAudio WaveProvider32 etc. When this audio output requests samples to be played, alphaTab grabs them from an internal audio buffer. If this audio buffer is filled less than 50%, it requests another 50% of the audio samples. Internally a loop starts which sends midi events to TinySoundFont and generates the audio samples.

But to connect an appropriate midi output, you need a real time sequencer, which plays the midi events in the exact speed. Today you would always receive in regular intervals, based on the internal audio buffers, a list of buffers which have been synthesized.

I don't know your exact use case yet. But in general there are two features related to the midi you might be interested:

1. You can generate the midi of a song independently using the low level APIs.
2. You can receive any played midi event in quasi-real time via a special event.

Depending on what you want to achieve, you might have the features already available. You can use any components understanding Midi and send the generated midi along. Some conversion between the event classes is obviously needed or you go via binary writing/reading. This can be any synth of midi library depending on your preference.

Or you use the original synth of alphaTab, but you don't load any SoundFont, but just use the internals to have a real time playback of "silence". Based on your needs you can forward the midi events or react on them manually to do something.

What I do not plan to support, is a separate real time sequencer which does a playback of the events in real time. There are enough libs for that.

[novaXire]

I didn't see the second link, i'll check this.

I just want to be able to use an external midi device or the built-in Microsoft GS Wavetable Synth or a virtual midi port to use a VST in a DAW for example in sync with the score.

[Danielku15]

I think in this scenario, you might be better set with a completely separate playback system independent from the default synthesizer. You are better set with a real time midi sequencer to send the output to any other synth or audio components.

There are various midi libraries you might use:
https://github.com/atsushieno/managed-midi
https://www.nuget.org/packages/Sanford.Multimedia.Midi/

Regarding the cursor display you would either need to use a complete own cursor overlay which works with your synth/playback system, or you might be able to call some low level apis like the normal synth:

alphaTab/src/AlphaTabApiBase.ts

Line 792 in f1f2af5

private cursorUpdateTick(tick: number, stop: boolean = false): void {

Theoretically the components within alphaTab are quite decoupled and can be used independent. But of course you might have to wire them up correctly to your execution path and logic.

Some parts of the cursor might be a bit tightly coupled with the built-in synth, but I guess we can easily work around or adapt this. Like: currently the cursors are animated according to the built-in player state and speeds.